The effect of formation processes on the frequency of palaeolithic cave sites in semiarid zones: Insights from Kazakhstan

Central Asian caves with Palaeolithic deposits are few, but they provide a rich record of human fossils and cultural assemblages that has been used to model Late Pleistocene hominin dispersals. However, previous research has not yet systematically evaluated the formation processes that influence the frequency of Palaeolithic cave sites in the region. To address this deficiency, we combined field survey and micromorphological analyses in the piedmont zone of south Kazakhstan. Here, we present our preliminary results focusing on selected sites of the Qaratau mountains. Sediment cover varies among the surveyed caves, and loess‐like sediments dominate the cave sequences. The preservation of cave deposits is influenced by reworking of cave sediments within the caves but also by the broader erosional processes that shape semiarid landscapes. Ultimately, deposits of potentially Pleistocene age are scarce. Our study provides new data in the geoarchaeologically neglected region of Central Asia and demonstrates that micromorphology has great analytical potential even within the limitations of rigorous survey projects. We outline some of the processes that influence the formation and preservation of cave deposits in Kazakhstan, as well as broader implications for the distribution of Palaeolithic cave sites in Central Asia and other semiarid environments.


| INTRODUCTION
Within the approximately four million square kilometres that span the five Central Asian Republics of Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan, only 18 caves document Palaeolithic occupation (see Figure 1). These sites are located in the intermontane basins and river valleys that shape the foothills of the high-altitude Central Asian mountain massifs. The Russian Altai, located at the northern fringes of Central Asia, have the highest frequency of Palaeolithic cave sites in the region, with a geographically restricted cluster found along the tributaries of major rivers. Further south, isolated Palaeolithic cave sites have been found in Kazakhstan, Kyrgyzstan and Tajikistan, while the second cluster of sites is reported along the Alay mountains in Uzbekistan. The Palaeolithic occupation of Central Asian caves ranges from the Middle to the Upper Palaeolithic, and despite their low numbers, in many cases, they have provided rich cultural assemblages and human remains (see Table S1). Analysis of these palaeoanthropological remains has led to novel genetic discoveries regarding human evolution, such as the identification of the Denisovan hominin group Reich et al., 2010;Slon et al., 2018). Building upon this record and in combination with data from open-air sites, various studies have attempted to model the presence of hominins in the Central Asian landscape (Beeton et al., 2014;Glantz et al., 2018;Iovita et al., 2020;Li et al., 2019). It seems that the foothills that connect the Central Asian mountains towards the West and the desert/steppe zones towards the East form an Inner Asian Mountain Corridor (IAMC; Frachetti, 2012) that may have served as a likely location of hominin refugia (Beeton et al., 2014;Glantz et al., 2018). Especially during glacial conditions, a 'northern' route along the foothills of the IAMC appears as the sole most likely scenario for hominin dispersal across Central Asia (Iovita et al., 2020;Li et al., 2019).
Even though these models provide important implications regarding the distribution of Palaeolithic sites in Central Asia, their accuracy is limited by the quality and quantity of the available data set. In particular, the Russian Altai is the only well-studied area in the region, being the subject of multidisciplinary research since the 1980s (Derevianko et al., 2018, p. 303). However, survey and excavation projects have been fewer south of the Altai, where the relative absence of systematic survey may have implications for the low distribution of cave sites (Fitzsimmons et al., 2017). We know little about the formation processes of the archaeological record in this region, since a high-resolution contextual methodology has been applied only on selected sites associated with hominin remains. In those cases, geoarchaeological approaches using a microanalytical methodology (Mallol et al., 2009;Morley et al., 2019) or broad-scale observations Krivoshapkin et al., 2020) have significantly aided our understanding of geogenic deposition, anthropogenic impact and local environmental change. These studies have broader archaeological importance since the analysis of cave sediments in arid to semiarid environments, like Central Asia, is rather limited. In this context, we stress that our picture for Late Pleistocene Central Asia is made up of only a few individual well-studied cases, extrapolated models and limited knowledge of the processes that govern the archaeological record on a regional scale. To change this picture, we require more field data to help us understand how the interaction between hominins and geomorphic environments shaped the unique Late Pleistocene archaeological record along the IAMC. In our recent paper (Iovita et al., 2020), we presented preliminary results of the 2017-2019 survey in Kazakhstan and attempted to evaluate some taphonomic biases that influence the distribution and quality of archaeological sites in the region. Here, we build further upon that study to explore the occurrence and characteristics of cave sediments in South Kazakhstan. First, we present statistics on the presence of sediment in caves and rockshelters based on the total number of features surveyed and test-excavated by our team. To assess the completeness of our data set, we utilise observations on cave morphology to examine the potential erosion of pre-existing sediments. Second, we focus on the Qaratau mountains and combine field stratigraphy with micromorphology to explore the depositional processes operating at different cave sites within that range.

| The Qaratau mountains in the context of the Inner Asian Mountain Corridor (IAMC): Geographic setting and geology
The IAMC constitutes a 2500 km-long chain of mountain foothills (piedmonts) flanked by lowland deserts (e.g., Qyzylqum Qaraqum Moyunqum, Tauqum, Saryyesik-Atyrau) and high mountains (the Pamir, Alay, Tian Shan, Dzungar and Altai), extending from Afghanistan to southern Siberia (see fig. 1 in Iovita et al., 2020). The majority of stratified Palaeolithic sites in Central Asia are found in this piedmont zone, which appears to have functioned as an ecological niche fostering hominin dispersals (Beeton et al., 2014;Glantz et al., 2018;Li et al., 2019;Zwyns et al., 2019). About half of the area of the IAMC falls within the modern territory of Kazakhstan, whose Palaeolithic settlement patterns remain relatively understudied (Cuthbertson et al., 2021). Complex and tectonically active landscapes, such as the Kazakh piedmonts, would be attractive for Palaeolithic hunter-gatherers since they provide availability of water, shelter and rich animal and plant resources in contrast to the desert and steppe lowlands that dominate the regional topography (Bailey & King, 2011;Winder et al., 2015). However, the Kazakh piedmonts could also be attractive for archaeologists since they preserve archaeological sites in different geomorphic contexts such as caves, loess-mantled slopes and springs (Iovita et al., 2020). Loess sediments dominate the Quaternary deposits in this piedmont zone, providing both a potential sediment source for the formation of archaeological sites and a palaeoenvironmental archive (Fitzsimmons et al., 2017).
By conducting a thorough survey of carbonates in four distinct regions of the Kazakh piedmont, our team concluded that the majority of surveyed caves, including the caves presented in this study, are found in the Qaratau mountains (Cuthbertson et al., 2021;Iovita et al., 2020).
The Qaratau mountain range is located in South Kazakhstan, delimited by the Qyzylqum desert, the Syr Darya and Arys rivers to the West, the Chu-Sarysu basin and Moyunqum desert to the East, the South Turgay basin to the North and the Tian Shan Mountains to the South (Figure 2). It has a NW-SE trend and is divided into two ridges: the Lesser Qaratau in the southeast and the Greater Qaratau in the northwest. Overall, the Qaratau mountains constitute a Northern segment of the major Talas-Fergana fault (Alexeiev et al., 2017;Burtman, 1980), with their evolution tied to the broader patterns of Central Asian tectonics (e.g., Kirscher et al., 2013).
Some of the oldest and most abundant rock types found in the Qaratau mountains include siliciclastic and volcanic rocks of Neoproterozoic age, as well as Middle and Upper Ordovician marine carbonates and granitoids. Towards the Middle Palaeozoic, volcanism and sedimentation in the region were generally associated with the passive margin development that contributed to the progressive amalgamation of the Palaeo-Kazakhstan continent (Biske, 2015). Regarding these changes, the formation of a carbonate platform from the Late Devonian until the Middle Carboniferous testifies to the presence of the Turkestan Ocean in the vicinity of the Qaratau and marks a new period of carbonate deposition in the area. This carbonate sequence is about 4 km thick, outcrops frequently throughout the mountain range and consists of depositional facies with diverse lithology (Cook et al., 2002). The geological picture of the area changed drastically after the Late Carboniferous, when major deformation events led to marine regression, termination of carbonate sedimentation and uplift (Alexeiev et al., 2009). Continental accretion culminated during the Late Palaeozoic, resulting in the closure of the Palaeo-Asian ocean and the formation of the Central Asian Orogenic Belt (Windley et al., 2007). Successive reactivations of the Talas-Fergana fault during the Mesozoic and Cenozoic induced additional deformation in the Qaratau. In the Jurassic, an elongated depression (Leontiev Graben) formed between the Greater and Lesser Qaratau, accumulating coal-bearing lacustrine and fluvial sediments (Alexeiev et al., 2017;Allen et al., 2001). In the Cenozoic, the collision between India and Eurasia about 50-35 Ma initiated substantial orogeny, with modern Tien Shan relief developing after~3 Ma (Buslov et al., 2008;Trifonov et al., 2008). The interplay between Quaternary climatic evolution and local neotectonics dramatically changed the environments of East Kazakhstan. Glaciations and increased aridification led to extensive deposition of glacial and aeolian sediments covering intermontane basins and their adjacent foothills (Aubekerov, 1993;Chlachula, 2010).
In contrast to other parts of Central Asia and Kazakhstan, the major uplift in the Qaratau enables the exposure of pre-Cenozoic structures that would otherwise be masked by recent sediments (Allen et al., 2001, p. 84). This setting facilitates the survey of the karst-forming Palaeozoic carbonate sequence and provides implications for the clustering of caves and rockshelters in this part of Kazakhstan. The limited speleological work in the region demonstrated that cave formation in some parts of the Qaratau is associated with Carboniferous karst massifs and plateaus shaped by tectonics (Shakalov, 2010(Shakalov, , 2011.
As a survey project, we decided against this high-resolution approach since (1) we aimed for a broad investigation of caves and rockshelters in our survey area, rather than focusing on a long campaign of excavating a single site, and (2) we could not apply an exhaustive range of analytical techniques because of logistical constraints on time in the field, as well as transport and storage during long survey campaigns. Instead, we used micromorphology selectively to gain a plethora of contextual information within promising sites, to interrogate difficult stratigraphic relationships and to establish a connection between landscape and site-specific processes. While the micromorphological results presented here are not exhaustive and do not aim to reconstruct the whole range of formation processes operating at a given site, they provide preliminary insights into the characteristics of the excavated sequences by highlighting the dominant depositional factors that operate at these different localities.

| Survey methodology
The caves and rockshelters presented here were surveyed and recorded during our recent fieldwork in Kazakhstan (Iovita et al., 2020). The surveys were structured around a novel modelled approach (Cuthbertson et al., 2021)  and high slopes that bound deep valleys (Cuthbertson et al., 2021).
For the on-site recording of features, we used an adapted version of the PaleoCore data structure (PaleoCore.org; Reed et al., 2015Reed et al., , 2018, and focused primarily on morphological attributes that were likely to be useful for further archaeological and geological investigations (e.g., sediment presence, cave morphology, speleothems). Most of the caves in our study area are single-chambered caves, and their formation history appears to be closely related to tectonics. For more information on cave and rockshelter morphology in our study area, see Iovita et al. (2020).

| Sediment occurrence, stratigraphic documentation and micromorphology
A primary goal of our survey was to test the archaeological potential of caves in Kazakhstan. We used sediment thickness in individual caves as a guide to focus on prominent sites, based on the assumption that thicker cave sequences would have higher chances of preserving archaeological deposits or Pleistocene sediments. The influence of modern cave use in the formation of the archaeological record has not been documented in Kazakhstan in the past. However, ongoing ethnographic work by our team demonstrates that caves are mostly associated with religious practices that do not heavily rework the deposited sediments (Bigozhin et al., unpublished data). Because caves are rarely used for pastoral activities like stock-keeping, distinct stabling deposits, which are common in other parts of the world (e.g., Angelucci et al., 2009), were not found during field survey. Reworking of older cave sediments is documented only at the site of Tuttybulaq 1, induced by smelting activities dating to the medieval period (Baytanaev et al., 2017(Baytanaev et al., , 2018(Baytanaev et al., , 2020. Therefore, by documenting sediment characteristics across different caves, we built a regional data set of cave sediment distribution that serves as a basis for exploring the depositional and erosional processes that influence the formation of the cave record. To explore the potential erosion of pre-existing sediments in empty caves, we focused on the recording of specific morphological characteristics that could indicate erosional events in the interior and the exterior of karst features. Regarding the interior of karst features, we searched for past cave surface levels, remnant sediment pockets (unconsolidated or cemented) and evidence for the differential weathering of cave wall surfaces induced by sediment removal (O'Connor et al., 2017). Turning to the exterior of karst features, we investigated the adjacent topography to identify rockfall and debris accumulations (e.g., talus slopes) that could be associated with largescale erosion of the features themselves.
For caves with sediment, we classified sediment thickness in both unexcavated and excavated features. In unexcavated features, we estimated sediment thickness as a minimum value from field observations of cave morphology, and where possible, we used a dynamic cone penetrometer (Kessler Soils Engineering, Inc.; Model K100) to verify our assessments. For excavated caves, we documented sediment thickness based on older publications or from our new test trenches. Our classification scheme was heuristic and used three levels of sediment cover: caves with 'Minor' deposits (<0.5 m),

| Thin-section preparation procedure and analysis
The micromorphology samples were encased in plaster, and after extraction, were wrapped with paper and packaging tape to ensure integrity during transport. Thin sections were produced in the Geoarchaeology Laboratory at the University of Tübingen and Terrascope Thin Section Slides. Initially, the samples were dried in an oven at 40°C and impregnated with a mixture of polyester resin, styrene and methylethylketone peroxide (MEKP) hardener under vacuum. After a period of around 20 days, the block samples reached the required hardness and were sliced into slabs with a rock saw. The thin-section production procedure ended with the mounting of the slabs onto 6 × 9 cm glass slides, and then grinding of these slabs to about 30 μm thickness. For some samples, a third mounting or hand polishing was necessary to obtain the right thickness. The thin sections were initially scanned using a high-resolution flatbed scanner to be documented and examined macroscopically (Haaland et al., 2019). Afterwards, they were studied under a stereoscope (0.65 -5× magnification) as well as a petrographic microscope (20-500× magnification) using plane-polarised light (PPL), cross-polarised light (XPL) and oblique incident light. Micromorphological descriptions follow the nomenclature and criteria proposed by Stoops (2003) and Courty et al. (1989) and are presented in Table S4. Thin sections were also examined under a fluorescent microscope equipped with the Zeiss Colibri system by using the 470 nm filter to test for phosphate and the 555 nm filter to test for organics.

| RESULTS AND INTERPRETATIONS
During the fieldwork seasons of 2017-2019, we surveyed a total of 95 caves and rockshelters (Table S2). Sixty-seven features are devoid of sediment and 28 have a varying degree of sediment cover. Out of the 28 features, eight caves had already been excavated in the past; we conducted test-excavations in 10 caves in total, including five newly documented caves (Table 1) (Table 1).   Here, we present our observations from the field and results of micromorphological analysis from five caves of the Qaratau mountains (Jetiotau, Qyzyljartas, Ushozen 1, Qaraungir 1 and Aqtogai 1; Figure 2, Table 1, Figure S1). We selected these five caves since their diverse sequences provide an overview of the major processes that seem to influence the formation of cave sites in the region.

| Jetiotau
The Jetiotau cave is located~2 km north-east from the Janatalap village of the Baidibek district, Turkestan region. It is formed on Lower Carboniferous (Tournaisian) carbonates at the South Western part of the lesser Qaratau, adjacent to the fault zone forming the Leontiev graben. It has a NW-SE orientation and a tube-shaped morphology consisting of a single 30 m-long passage with a maximum roof height of~7 m (see also Figure S2a).

| Stratigraphic overview
In Jetiotau, we excavated a 3 × 1 m test trench at the entrance area of the cave, exposing a stratigraphic sequence of 2.12 m without reaching bedrock (Table S3) This deposit mainly consists of micrite with the addition of wellsorted silt, sand-sized quartz and mica grains in the coarser laminae.
The fluctuating composition of the laminae is indicative of sheetwash processes (Karkanas & Goldberg, 2018), while the parallel to subparallel orientation of mica grains (see Figure 5b) also suggests deposition in a low-energy water-lain environment (Mücher & Ploey, 1977). In comparison with MU J3-1, MU J2-1 also contains charcoal fragments and has a more granular microstructure. Overall, the micromass of MU J3-1 appears to be more phosphatic and isotropic in XPL. In places, the phosphatisation is accompanied by de-calcification, judging from the absence of a crystallitic bfabric and the removal of calcite in altered limestone clasts.
However, in contrast to this decalcified matrix, we observed many calcitic-crystallitic aggregates as well as bone fragments heavily cemented by calcite (Figure 5d). The considerable variation in postdepositional processes (decalcified vs. calcified components) in the same deposit is a strong indication that MU J3-1 represents a mixture of different sediment sources.
MU J1-1 is a moderately sorted deposit with sand-sized charcoal and bone fragments that comprises the uppermost part of the sequence, corresponding to LU J1 and modern cave use. It has a similar fabric to MU J2-1. The granular microstructure at the top part of LU J1 and the high frequency of channel voids demonstrate extensive bioturbation.

| Qyzyljartas
The Qyzyljartas cave is located at the north-eastern foothills of the Greater Qaratau range, about 10 km south-west of the Sozaq town. It is formed at the top of a steep sandstone outcrop (Figure 15c), while the feature itself has three openings, two of which join together to create a long, funnel-like cave, open at two sides. Two sloped passageways are oriented southwest and south (see also Figure S2c).  Iovita et al., 2020, p. 123). has higher abundance of interstitial clay.

| Ushozen 1
Ushozen 1 is a cave located ca. 10 km northwest of the Babaiqorgan village, Turkestan region, on the eastern bank of the homonymous Ushozen river. It is formed on Lower Devonian carbonates of the Aman formation at the northwestern part of the Greater Qaratau.
The cave is composed of a single chamber, approximately 7 × 8 m (see also Figure S2d).

| Aqtogai 1
The Aqtogai 1 cave lies on the right bank of the Shabaqty river, about 10 km southeast of the Janatas town, Jambyl region, at the eastern part of the Lesser Qaratau (see also Figure S2e). It is formed on Middle Ordovician limestone, at an uplifted and highly deformed mountain front bounded by the Greater Qaratau Fault structure (Allen et al., 2001, p. 89).

| Stratigraphy overview
In Aqtogai 1, we expanded a test trench (

| Micromorphology
Micromorphology sample PSR-19-6 is classified into two MUs (A6 and A7) corresponding to the contact between LUs A6 and A7. MU A3 is a heterogeneous organic-rich deposit corresponding to LU A3. It consists of numerous rock fragments, phosphatic grains, endokarstic silty clay clasts and dung pellets (Figures 11e and 14c).
Dung shows a varying degree of preservation like in MUs A6 and A7.
The coarse material shows uniform dipping and orientation and is occasionally microlayered (Figure 11e). We hypothesise that the preferential arrangement of coarse components and the microlayer-

| Qaraungir 1
Qaraungir 1 cave is located in the foothills of the Lesser Qaratau range, 30 km northeast of Shymkent in southern Kazakhstan. The inner part of the cave has been previously excavated by Taimagambetov and Nokhrina (1998), with the oldest deposits dated to the Neolithic (see also Figure S2b).

| Stratigraphy overview
Building upon the work of Taimagambetov and Nokhrina (1998), who documented Neolithic occupation in the interior of the cave, we decided to excavate outside of the dripline to assess the lateral distribution of archaeological deposits. Our test trench at Qaraungir 1 reached a maximum depth of~140 cm, exposing scarce Holocene archaeological material throughout the sequence, but no dense cultural layers ( Figure 12). Therefore, the work of Taimagambetov and Nokhrina (1998) and our investigations suggest that the sediments inside and outside of the dripline in Qaraungir 1 were most probably deposited during the Holocene. The LUs have a silty clay to clayey loam texture, with a high frequency of coarse clasts especially in LU QA3 and towards the bottom of the trench. The shallow stratigraphy and the absence of cultural layers contrast with the thick cultural sequences recorded inside the cave by Taimagambetov and Nokhrina (1998). Therefore, Qaraungir 1 is the only surveyed cave where we have enough data to explore spatially diverse formation processes. Additionally, Qaraungir 1 is one of the few caves located in a down-slope position, providing an opportunity to study processes that may not be active in caves located in areas of higher topographic relief. Pending OSL dates will demonstrate when the sediments were deposited in the slope of Qaraungir 1.

| Micromorphology
MU QA3 and QA2 are both clast-supported deposits that consist of various geogenic and biogenic components (Figure 13). Although

| DISCUSSION
Our survey and micromorphological data suggest that the accumulation and preservation of sediments vary among the Qaratau caves.
Below, we present a discussion of the processes that influence the F I G U R E 11 Microphotographs from Aqtogai 1 cave. (a) MU A7; randomly distributed coarse-sized limestone fragments (lm) and silty clay clasts (sc) mixed with dung pellets (dp), degraded dung (arrows) and phosphatised material (ph). Cemented deposits (cd) and a speleothem fragment (sp) are also present, indicating the mixing of heterogeneous deposits; PPL. (b) MU A5; gravel-sized dung pellets (dp) and few silty clay clasts (sc) embedded in an ashy matrix; XPL. (c) MU A4; gravel-sized and comminuted charcoal (ch), sediment aggregates (sa) and common isotropic phosphatic aggregates ( | 607 distribution of cave sediments in respect to the regional semiarid context.

| Summary of site formation processes in the Qaratau caves
Aeolian input leads to the formation of loess-like cave sediments that share common mаcroscopic characteristics across the cave sites. These sediments can be identified in the field based on pale colour, silty texture and massive structure (see also Krajcarz et al., 2016). Based on our micromorphology analysis, we assume that these textural attributes result from similarities in the micromass, which is characterised by the high abundance of very fine sand to silt-sized quartz, mica grains and calcite. However, under the microscope, loess-like cave sediments also demonstrate a high degree of compositional variability, as they mix with a wide range of materials depending on the cave environment. Therefore, homogeneous wind-blown loess deposits were not observed in any of the caves, suggesting that the loess-like material found within the caves was likely reworked through a number of different processes. In contrast to grain shape, in this study, we demonstrated that grain orientation constitutes an especially useful tool for identifying postdepositional processes of loess-like cave sediments. Under the microscope, uniformly oriented mica particles may constitute a proxy of water reworking, or even form deformation features in a mass movement context. However, due to the homogeneity of the loess matrix, low-energy reworking cannot always be observed in the micromass. Therefore, we suggest that the distribution and depositional history of the coarser sand-sized material that becomes mixed with loess is usually more helpful in documenting reworking in loesslike cave sediments.
Based on our survey results, the majority of the examined caves are hydrologically abandoned in the sense that they are decoupled from any major groundwater input (Sherwood & Goldberg, 2001). As demonstrate more recent water-driven processes. In this context, the frequent occurrence of low-energy colluvial (Qaraungir 1, Aqtogai 1) or higher-energy mass movement processes (Jetiotau) near the cave entrance also requires some degree of water saturation (Karkanas & Goldberg, 2018). We hypothesise that regional orographic precipitation supplies the necessary water content driving the depositional processes described above, which may occasionally trigger a reactivation of the karst network. Because of higher relief, the Qaratau mountains and the greater Tian Shan mountain range are characterised by higher mean annual precipitation values and more frequent precipitation extreme events in comparison to other regions of Central Asia (Ma et al., 2020).
The depositional processes outlined above have diverse implications for the preservation of cave sequences. First of all, the thick aeolian deposits demonstrate that there are extensive periods of time where stable conditions without groundwater flow enabled the settling of loess into the caves. Cave surfaces must have been exposed for a significant amount of time based also on the high content of phosphatised and calcified material (Barbieri et al., 2018;Miller, 2015). Except from phosphatisation, diagenetic processes are mainly linked to the formation of authigenic gypsum in Aqtogai 1, indicating mostly dry conditions. The absence of intensive diagenetic processes demonstrates that the Qaratau caves show good potential for the preservation of organic materials. In this regard, the case study from Aqtogai 1 demonstrates that the high frequency of organic materials is of high importance for the build-up of thick cave sequences.
4.2 | Investigating cave erosion by combining field survey and micromorphology Sherwood and Goldberg (2001) suggested that postdepositional alteration of cave sediments is site-specific, as it is controlled by microenvironmental factors such as bedrock characteristics, landscape location, local hydrology and human activity. Despite site variation, our field survey and micromorphology work in the Qaratau mountains revealed that regional patterns of sediment preservation and reworking may be inferred.
Understanding the processes that accumulate or remove cave sediments in Kazakhstan is a major challenge since most of the surveyed caves and rockshelters did not contain any sediments. In features. In addition, evidence for ongoing sediment erosion is also minimal. Active erosional processes were recorded only in Nazugum rockshelter (Iovita et al., 2020), where we documented water channels washing out parts of the sequence (Figure 15b).
Generally, traces of erosion are more frequently related to processes affecting the exterior of karst features. In Qaratau, semiarid conditions hinder the development of thick soils, facilitating the formation of scree-mantled slopes and talus cones (Abrahams et al., 1994). Based on the high frequency of these erosional landforms in the mountain foothills of the surveyed areas, we hypothesise that caves or cave sediments might have been eroded from the landscape. In this context, the caves and rockshelters that we surveyed are usually found in a mid-slope position (Cuthbertson et al., 2021), overlooking these erosional scree slopes (e.g., towards the slope outside of the dripline is an indication that colluvial processes also influence the preservation of deposits in the few caves that are associated with soil-mantled slopes. Overall, the scarcity of Pleistocene sediments in contrast to the more common Holocene sediments (Iovita et al., 2020) could indicate that erosional processes affecting cave deposits were more intense during the Pleistocene. Even though this study demonstrated some potential pathways of cave erosion in specific sites, at this stage, we cannot provide a more detailed chronological framework for the onset of erosional processes for the whole range of the Qaratau mountains. Future work in prospective sites and their corresponding catchments will address the probability, intensity and chronology of erosion.

| The Qaratau caves in the context of Central Asian Palaeolithic and semiarid zones
Our survey in the Qaratau mountains has significant implications for the formation of the archaeological cave record in Central Asia (see also Iovita et al., 2020). Despite the numerous caves that we recorded during our survey, only a few contain thick sediment sequences. A similar situation seems to occur in Uzbekistan and neighbouring Mongolia, where recent surveys recorded only a few cave sites (Nishiaki et al., 2018(Nishiaki et al., , 2019Vanwezer et al., 2021). The formation of cave sites requires human activity and a geomorphological setting that promotes the accumulation and preservation of sediments (Mentzer, 2017). The geological structure is important for the preservation of sediments, and cave sites formed in rock strata that slope downwards tend to be eroded away under long time scales (Heydari, 2007). Besides rock type and structure, climate is the other major influence on the type of sediments deposited in a landscape and the pathways of its subsequent erosion (Bull, 2009;Burbank & Pinter, 1999;Ke & Zhang, 2021). However, the impact of climate on the evolution of cave sediments has been contextualised only for some environments in the geoarchaeological literature, such as Mediterranean and tropical (Morley, 2017;Woodward & Goldberg, 2001). Central Asia and other arid or semiarid settings have been largely neglected in the discussion of cave-formation processes, probably due to the lack of a group of   Li et al., 2015Li et al., , 2020. The variability in the distribution of aeolian loess sediments, together with the erosional processes presented in this study, could potentially explain the frequency of empty caves in Qaratau mountains. The limited soil cover in semiarid areas (e.g., Figure 15c,d) also hinders the redeposition of soil material in the caves through colluvial processes.
This type of allochthonous colluvial sediment is important for the build-up of cave sequences in slightly more humid climates, such as dry-Mediterranean (Frumkin et al., 2016;Woodward & Goldberg, 2001). Nevertheless, the alteration of hot and cool conditions that are also present in semiarid areas facilitates the thermostatic weathering of the bedrock and leads to the accumulation of angular limestone debris in cave sequences (Cremaschi et al., 2015). Roof spall and remobilised karstic sediments constitute the dominant autochthonous geogenic deposit that we recorded in our survey. In the case of pseudokarstic caves, such as Qyzyljartas, the disintegration of non-carbonate bedrock into loose sediment will provide an extra source of autochthonous sediment accumulation (see also Iovita et al., 2020). These autochthonous deposits mix with the aeolian component by colluvial and mass movement processes triggered inside the cave environment. Other processes, such as spring activity and sheetflow processes, have only been recorded at Obi-Rakhmat (Mallol et al., 2009), and we hypothesise that they are relatively rare in Central Asian and semiarid caves, since we also recorded them only in rare instances (e.g., Qyzyljartas and Nazugum).
The alteration of aeolian deposition and geogenic colluvial reworking seems to be a recurring pattern not only in caves of the semiarid part of Central Asia (this study and Sel'ungur; Krivoshapkin et al., 2020) but also in the caves from the boreal and more humid Altai region. Available data from Strashnaya (Krivoshapkin et al., 2018(Krivoshapkin et al., , 2019, Chagyrskaya  and Ust'-Kanskaya (Lesage et al., 2020) suggest that some cave sequences in the Altai are punctuated by the accumulation of loess-like sediments and autochthonous colluvial reworking. However, Altai caves are also often characterised by cryogenic deformation features, most probably induced by the more boreal and humid local climatic conditions Krivoshapkin et al., 2019;Morley, 2017).
These features are postdepositional and constitute an additional agent of sediment mixing. In contrast, cryoturbation features have not yet been reported in the more arid southern Central Asia, which could imply less intense postdepositional processes and more secure cave contexts.
Despite the more intense postdepositional processes, the Altai region has a much higher frequency of Palaeolithic cave sites in comparison to Central Asia. If we adopt a 'simplistic' climatic approach to the data, we could argue that the distribution of cave sites reflects solely different climatic conditions. According to this approach, the Altai cluster reflects a more diachronic occupation favoured by the overall better climatic conditions, while semiarid Central Asia functions only as a corridor that witnesses substantial occupation only during phases of ameliorating climate. This approach, however, would not be valid based on the recent modelling data that suggest the presence and movement of hominin groups in the IAMC during both glacial and interglacial conditions (Glantz et al., 2018;Li et al., 2019). While the reasons for this preferential distribution of cave sites remain unclear, we believe that they also reflect variations in the processes that influence the formation of cave sediments and the stability of caves on the landscape. More evidence on regional site formation processes would greatly enhance the challenging task of correlating site distribution with human choice and dispersal routes.

| Methodological implications
In this study, we demonstrated that micromorphological analysis could provide valuable information in archaeological surveys. By collecting qualitative data from several sites, we answered questions that often remain unaddressed by survey projects that focus primarily on the quantitative distribution of sites on the landscape. The occurrence and thickness of sediment cover, the origin of cave deposits, depositional processes and postdepositional alterations are key site-specific parameters that could not have been explored using a purely landscape approach. Incorporating this information together allows us to examine the dominant processes that control the formation of the record but also demonstrates the degree of variation within a specific region. In the Qaratau example, we have demonstrated that even though loess is the main driver of allochthonous sediment accumulation, the way it gets reworked among the different caves varies greatly. In this regard, formation processes are not only influenced by site location but also by the sitespecific depositional history. Other processes, such as anthropogenic input (e.g., at Aqtogai 1), or rare depositional processes (e.g., at Qyzyljartas) could form cave sequences that stand out from the rest of the data set. Moreover, by combining macroscopic observations for the whole data set together with site-specific analysis, we were able to address how representative our interpretations are in a broader sense. In this way, we supply the reader with data that are often omitted in archaeological survey publications. Even for sites of low archaeological potential, our micromorphological survey approach enables us to reconstruct cave life histories and model the potential formation processes that characterise our study area (see also Karkanas et al., 2021) and also to potentially examine factors of human absence in the landscape as well as presence.

| CONCLUSIONS
This study provides a preliminary geoarchaeological context for our ongoing cave survey in the Qaratau mountains of South Kazakhstan (Iovita et al., 2020). By combining model-led intensive field survey (Cuthbertson et al., 2021) with micromorphological analysis, we assessed the distribution of cave sediments and prominent caves on the landscape and demonstrated how cave-formation processes are tied to the regional geomorphological and climatic factors. This study has implications for caves in similar semiarid settings and provides a methodology for contextualising survey data with a high-resolution analytical framework. Thus, it addresses themes that often remain unaddressed in the (geo) archaeological literature since welldocumented semiarid caves sites are lacking, fieldwork projects often do not carry out high-resolution site-specific analyses and micromorphology studies often do not utilise a regional approach by focusing on a group of different cave sites.
Qaratau caves recorded different depositional styles, but loess-like cave deposits and reworking processes of varying intensity dominate the sediment sequences. Moreover, the depositional and erosional processes that characterise the surveyed caves are also associated with their landscape location. We hypothesise that hillslope erosion might influence the removal of caves from the landscape, and in combination with loess cover, might blanket caves found downslope.
Overall, a new Denisova-type cave has not yet been found during our survey in the Qaratau mountains. Caves with the potential for Pleistocene sediments were inferred only from a couple of sites, and future excavation and dating are required to resolve the sedimentary record of these caves. To date, only two Palaeolithic cave sites are known from Kazakhstan, even though the number of Palaeolithic open-air sites is gradually increasing (Anoikin et al., 2019;Ozherelyev et al., 2019). However, the low frequency of Palaeolithic cave sites is a general characteristic of the caves found in the semi-arid regions of Central Asia and contrasts with the high clustering of Palaeolithic cave sites found in the more humid northern fringes of the Altai. This distribution cannot be explained only by climatic factors, and in this paper, we present some of the formation processes that influence the deposition and erosion of sediments in Central Asia. We hypothesise that additional geological factors such as the distribution and type of karst landscapes, together with the subsistence strategies used by hominin groups in semiarid environments, shape the complex Central Asian Palaeolithic record. A methodology focusing on survey and high-resolution analysis, similar to the one used in this study, has the potential to unravel this record and provide the necessary data for further modelling research targeting human dispersals in the region.